Notes on Dierential Forms

Ivan Avramidi
New Mexico Institute of Mining and Technology
Socorro, NM 87801
December 2003
Differential Forms 1
1 Tensor Algebra
1.1 Manifolds and Local Coordinates
Let M be an n-dimensional smooth orientable manifold without boundary.
Then locally, at any point x
0
M, there is a neighborhood such that it
can be dieomorphically mapped to a region in the Euclidean n-dimensional
space R
n
with the coordinates x

, where = 1, . . . , n. What follows is a list

of useful formulas in that local coordinate chart with these local coordinates.
1.2 Tangent and Cotangent Spaces
The tangent space T
x
0
M at the point x
0
is a vector space spanned by the
basis
e

= /x

(1)
(coordinate basis). A tangent vector v can be represented by a n-tuple v

,
i.e.
v = v

. (2)
The cotangent space T

x
0
M at the point x
0
is a vector space of linear maps
: T
x
0
M R, v , v) , (3)
spanned by the basis

= dx

(4)
(coordinate basis). This basis is dual to the basis e

in the sense that

, e

) =

. (5)
A cotangent vector can be represented by a n-tuple

; then
=

(6)
and
, v) =

. (7)
(Recall that a summation over repeated indices is performed.)
Differential Forms 2
1.3 Tensors of Type (p, q)
A tensor of type (p, q) is a real valued multilinear map
A : T

x
0
M T

x
0
M
. .
p
T
x
0
M T
x
0
M
. .
q
R. (8)
A basis in the vector space of tensors of type (p, q) can be dened by
e

1
e
p

1

q
. (9)
Then a tensor of the type (p, q) is represented by the components
A

1
...p

1
...q
, (10)
so that
A = A

1
...p

1
...q
e

1
e
p

1

q
. (11)
1.4 Riemannian Metric
A Riemannian metric is a symmetric tensor of the type (0, 2) whose compo-
nents g

are given by a symmetric nondegenerate positive denite matrix

g

. The Euclidean metric is given just by the Kronecker delta symbol, i.e.
g

=
_
1 if = ,
0 if ,= .
(12)
The Riemannian metric denes an inner product of vectors by
(v, w) = g

, (13)
and one-forms
(, ) = g

, (14)
where g

is the matrix inverse to the matrix g

. It establishes an isomor-
phism between the tangent vectors and the covectors (one-forms) by

= g

, v

= g

. (15)
Similarly, one denes the operation of raising and lowering indices of any
tensor of type (p, q).
Differential Forms 3
1.5 Dierential Forms
A tensor of type (0, s) is called skew-symmetric or (anti-symmetric) if it
changes sign when the order of any two of its arguments is reversed, i.e.

...
i
...
j
...
=
...
j
...
i
...
. (16)
The skew-symmetric tensors of type (0, p) (called p-forms or dierential
forms) form a subspace of
T

x
0
M T

x
0
M
. .
p
. (17)
For simplicity we will denote it by
p
.
Let S
p
be the permutation group of integers (1, . . . , p). The signature
sgn() (or sign) of a permutation =
_
1
(1)
...
...
p
(p)
_
S
p
is dened to be +1
if is even and 1 if is odd. Then for any p-form there holds

(1)
...
(p)
= sgn()

1
p
. (18)
Therefore, a p-form is given by its components

1
p
where
1
1
<
2
< <
p1
<
p
n. (19)
The other components are given by symmetry, and symmetry gives no rela-
tions among the components with increasing indices. From this it is evident
that the dimension of the space of p-forms in an n-dimensional manifold M
is
dim
p
=
_
n
p
_
(20)
for any 0 p n and is zero for any p > n. In other words,
p
= 0 if
p > n. In particular,
0
is one-diemsnional for p = 0 and p = n.
1.6 Exterior Product
For any tensor T of type (0, p) we dene the alternating (or anti-symmetrization)
operator Alt . In components the antisymmetrization will be denoted by
square brackets, i.e.
(Alt T)

1
p
= T
[
1
p]
=
1
p!

Sp
sgn()T

(1)

(p)
, (21)
Differential Forms 4
where the summation is taken over the p! permutations of (1, . . . , p).
Since the tensor product of two skew-symmetric tensors is not a skew-
symmetric tensor to dene the algebra of antisymmetric tensors we need
to dene the anti-symmetric tensor product called the exterior (or wedge)
product. If is an p-form and is an q-form then the wedge product of
and is an (p + q)-form dened by
=
(p + q)!
p!q!
Alt ( ) . (22)
In components
( )

1
...
p+q
=
(p + q)!
p!q!

[
1
...p

p+1
...
p+q
]
. (23)
The wedge product has the following properties
( ) = ( ) (associativity)
= (1)
deg()deg()
(anticommutativity)
( + ) = + (distributivity) ,
(24)
where deg() = p denotes the degree of an p-form .
A basis of the space
p
is

1

p
, (1
1
< <
p
n) . (25)
An p-form can be represented in one of the following ways
=

1
...p

1

p
=
1
p!

1
...p

1

p
=

1
<<p

1
...p

1

p
. (26)
The exterior product of a p-form and a q-form can be represented as
=
1
p!q!

[
1
...p

p+1
...
p+q
]

p+q
. (27)
Differential Forms 5
1.7 Volume Element
The n-form
=
1

n
(28)
is called the volume element. The components of the volume form denoted
by

1
...n
(29)
are given by so called completely anti-symmetric Levi-Civita symbol (or al-
ternating symbol)

1
...n
=
_
_
_
+1 if (
1
, . . . ,
n
) is an even permutation of (1, . . . , n),
1 if (
1
, . . . ,
n
) is an odd permutation of (1, . . . , n),
0 otherwise .
(30)
Furthermore, the space of n-forms
n
is one-dimensional. Therefore, any
n-form is represented as
= f
1

n
, (31)
with some scalar f. The n-form
_
[g[
1

n
, (32)
where
[g[ = det g

, (33)
and g

is the Riemannian metric, is called the Riemannian volume element.

1.8 Interior Product
The interior product of a vector v and a p-form is a (p1)-form i
v
dened
by
(i
v
)

1
...
p1
=
1
(p 1)!
v

1
...
p1
. (34)
One can prove the following useful formula for the interior product of a
vector v and the wedge product of a p-form and a q-form
i
v
( ) = (i
v
) + (1)
p
(i
v
) . (35)
Differential Forms 6
1.9 The Star Operator (Duality)
The star operator maps any p-form to a (n p)-form dened by
()

p+1
...n
=
1
p!

1
...p
p+1
...n
_
[g[g

1
g
pp

1
...p
. (36)
The operator satises an important identity: for any p-form there holds

2
= (1)
p(np)
. (37)
Notice that if n is odd then
2
= 1 for any p.
1.9.1 Examples (R
3
)
In the case of three-dimensional Euclidean space the metric is g

, the
bases of p-forms are:
1, dx, dy, dz, dx dy, dx dz, dy dz, dx dy dz . (38)
The star operator acts on this forms by
1 = dx dy dz, (39)
dx = dy dz, dy = dx dz, dz = dx dy, (40)
(dx dy) = dz, (dy dz) = dx, (dx dz) = dy, (41)
(dx dy dz) = 1 . (42)
So, any 2-form
=
12
dx dy +
13
dx dz +
23
dy dz (43)
is represented by the dual 1-form
=
12
dz
13
dy +
23
dx , (44)
that is
()

=
1
2

(45)
()
1
=
23
, ()
2
=
31
, ()
3
=
12
, (46)
Differential Forms 7
and any 3-form
=
123
dx dy dz (47)
is represented by the dual 0-form
=
1
3!

=
123
. (48)
Now, let and be two 1-forms
=
1
dx +
2
dy +
3
dz , =
1
dx +
2
dy +
3
dz . (49)
Then
=
1
dy dz +
2
dz dx +
3
dx dz (50)
and
= (
1

1
)dxdy +(
1

1
)dxdz +(
2

2
)dy dz ,
(51)
() = (
1

1
+
2

2
+
3

3
)dx dy dz . (52)
Therefore,
( ) = (
1

2

2

1
)dz (
1

3

3

1
)dy + (
2

3

3

2
)dx , (53)
[ ()] =
1

1
+
2

2
+
3

3
, (54)
or
( ) = , (55)
[ ()] = . (56)
2 Tensor Analysis
3 Exterior Derivative (Gradient)
The exterior derivative of a p-form is a (p + 1)-form with the components
(d)

1
...
p+1
= (p + 1)
[
1

2
...
p+1
]
=
p+1

q=1
(1)
q1

1
...
q1

q+1
...
p+1
. (57)
Differential Forms 8
It is a linear map satisfying the conditions:
d( ) = d + (1)
deg()
d , (58)
d
2
= 0 . (59)
For any n-form (a p-form with rank equal to the dimension of the manifold
p = n) the exterior derivative vanishes
d = 0 . (60)
One can prove the following important property of the exterior derivative
of the wedge product of a p-form and a q-form (product rule)
d( ) = (d) + (1)
p
(d) . (61)
3.1 Examples in R
3
Zero-Forms. For a 0-form f we have
(df)

f , (62)
so that
df = grad f . (63)
One-Forms. For a 1-form
=
1
dx +
2
dy +
3
dz (64)
we have
(d)

(65)
that is
d = (
1

1
)dxdy+(
2

2
)dydz +(
3

3
)dzdx . (66)
Therefore
(d)

, (67)
so that
d = (
2

3

3

2
)dx + (
3

1

1

3
)dy + (
1

2

2

1
)dz . (68)
We see that
d = curl . (69)
Differential Forms 9
Two-Forms. For a 2-form there holds
(d)

, (70)
or
d = (
1

23
+
2

31
+
3

12
)dx dy dz . (71)
Hence,
d =
1
2

=
1

23
+
2

31
+
3

12
. (72)
Now let be a 1-form
=
1
dx +
2
dy +
3
dz . (73)
Then
=
1
dy dz
2
dx dz +
3
dx dy , (74)
and
d = (
1

1
+
2

2
+
3

3
)dx dy dz , (75)
or
d =
1

1
+
2

2
+
3

3
. (76)
So,
d = div . (77)
3.2 Coderivative (Divergence)
Given a Riemannian metric g

we also dene the co-derivative of p-forms

by
=
1
d = (1)
pn+p+1
d . (78)
That is the coderivative of a p-form is the (p 1)-form
()

1
...
p1
=
1
(n p + 1)!

1
...
p1
p...n
_
[g[g
p
g

p+1

p+1
g
nn
(n p + 1)

_
1
p!

1
...p
p+1
...n
_
[g[g

1
g
pp

1
...p
_
(79)
It is easy to see that, since
2
= 1 and d
2
= 0, the coderivative has the
following property

2
= 0 . (80)
Differential Forms 10
From this denition, we can also see that, for any 0-form f (a function)
f is an n-form and, therefore, d f = 0i.e. a coderivative of any 0-form is
zero
f = 0 . (81)
For a 1-form , is a 0-form
=
1
_
[g[

_
_
[g[g

_
. (82)
More generally, one can prove that for a p-form
()

1
...
p1
= g

1
. . . g

p1

p1
1
_
[g[

_
_
[g[g

1
g

p1

p1

1
...
p1
_
.
(83)
4 Integration of Dierential Forms
Any dierential n-form can be integrated over the n-dimensional manifold
M. One needs to introduce an atlas of local charts with local coordinates
that cover the whole manifold. For simplicity, we will describe the integrals
over a single chart only. That is we have local coordinates x

that map a
region in the manifold M to a bounded region U in the Euclidean space R
n
.
This region is supposed to have some nice boundary U. The the integral
_
U
=
_
U

1...n
dx
1
dx
n
(84)
is just an ordinary multiple integral over the coordinates x
1
, . . . , x
n
, in the
usual notation _
U
=
_
U

1...n
(x) dx
1
dx
n
(85)
More generally, any dierential p-form can be integrated over a p-
dimensional submanifold N of an n-dimensional manifold M. Since N itself
is a manifold this case reduces to the case of integration of a n-form over
a n-diemsnional manifold. Clealy, it depends on the embedding of the sub-
manifold N in the manifold M. If x = (x

) = (x
1
, . . . , x
n
), = 1, . . . , n,
Differential Forms 11
are the local coordinates on the manifold M and u = (u
1
, . . . , u
m
) = (u
j
),
j = 1, . . . , p, are the local coordinates of the submanifold N, then
_
N
=
_
N

1
...p
(x(u))
x
[
1
u
1

x
p]
u
p
du
1
du
p
. (86)
The general Stokes Theorem states that for any smooth (n 1)-form
dened over a bounded region U of a n-dimensional manifold M (in partic-
ular, of R
n
) with a piecewise simple (no self-intersection) smooth boundary
U the following formula holds
_
U
d =
_
U
. (87)
Here it is assumed that the orientation of U is consistent with the orientation
of U. The same formula holds for orientable manifolds with boundary.
4.1 Examples
One-forms. If =

dx

is a one-form and U is a curve x

= x

(t),
a t b, then
_
U
=
_
b
a

(x(t))
dx

(t)
dt
dt . (88)
Two-forms. If =
1
2

dx

dx

is a two-form and U is a surface x

=
x

(u), u = (u
1
, u
2
) U, then
_
U
=
_
U
1
2

(x(u))J

(x(u)) du
1
du
2
, (89)
where
J

= e

1
e

2
e

1
e

2
, (90)
where e
1
and e
2
are tangent vectors to the surface dened by
e

j
=
x

u
j
. (91)
In three dimensional Euclidean space R
3
one can represent the 2-forms
and J by their duals. The dual to the 2-form J is a one-form
J = e
1
e
2
= n
_
[g[, (92)
Differential Forms 12
where n is the unit vector (normal to the surface since it is normal to both
vectors e
1
and e
2
), [g[ = det g
ij
and g
ij
is the induced Riemannian metric on
the surface dened as
3

1
(dx

)
2
= g
ij
(u)du
i
du
j
. (93)
Therefore, the above formula simplies to
_
U
=
_
U
() n
_
[g[du
1
du
2
. (94)